HLM dropped cases

[nschell]

Hello,

I am importing data from SPSS into HLM.  I have three levels of data.  I have no missing data, the data is sorted correctly, and all the IDs match up.  However, when I import the data, HLM drops 19/301 cases at level 2.  I imported levels 1&2 and 2&3 separately as two-level models to narrow down where the problem was, and it seems to be between levels 1&2, but I have no idea how to find out which cases it is dropping or why.  Does anyone have suggestions aside from missing data, mismatched IDs or sorting?  Does anyone know how to look at the data in HLM...just looking at the descriptives tells me I'm missing data but not which cases.  Or any suggestions on how to figure out which cases are being dropped?

Thank you,
Natalie

[lts1]

Hi Natalie,

    Yes, you can track down the dropped cases.  Unfortunately, it's a
somewhat painstaking and tedious process, and I do not know how to look at
data in HLM.

    That said, I would do is start with the statistics file created by HLM.
You can find them in a file called HLM3MDM.STS.  This is really a text file
that can easily be opened using notepad or your favorite word processor.  It
contains the descriptives for your variables separated by level.  I would
compare the descriptives from HLM to the ones in SPSS.  Once you identify
which variables differ, you can use the filter command to identify the cases
that are missing (select cases that fall out of the ranges listed in the MDM
statistics file).

    Also, what versions of HLM & SPSS are you using?

    I hope this helps.

        Best,
            Lisa



----- Original Message -----
From: "nschell" <[hidden email]>
To: <[hidden email]>
Sent: Monday, October 15, 2007 1:34 PM
Subject: HLM dropped cases

> Hello,
>
> I am importing data from SPSS into HLM.  I have three levels of data.  I
> have no missing data, the data is sorted correctly, and all the IDs match
> up.  However, when I import the data, HLM drops 19/301 cases at level 2.
> I
> imported levels 1&2 and 2&3 separately as two-level models to narrow down
> where the problem was, and it seems to be between levels 1&2, but I have
> no
> idea how to find out which cases it is dropping or why.  Does anyone have
> suggestions aside from missing data, mismatched IDs or sorting?  Does
> anyone
> know how to look at the data in HLM...just looking at the descriptives
> tells
> me I'm missing data but not which cases.  Or any suggestions on how to
> figure out which cases are being dropped?
>
> Thank you,
> Natalie
> --
> View this message in context:
> http://www.nabble.com/HLM-dropped-cases-tf4629052.html#a13217714
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
>
>>>> Error in line 4 of spssx-l.mailtpl: unknown formatting command <<<
> -> .................... <-
>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[nschell]

Hi Lisa,

You were right, that was completely painstaking, but quite helpful!  Thank you so much for the advice!  Now I know which cases are being dropped and I just need to figure out why...the fun never ends.

Many thanks,
Natalie

Lisa Stickney wrote

Hi Natalie,

    Yes, you can track down the dropped cases.  Unfortunately, it's a
somewhat painstaking and tedious process, and I do not know how to look at
data in HLM.

    That said, I would do is start with the statistics file created by HLM.
You can find them in a file called HLM3MDM.STS.  This is really a text file
that can easily be opened using notepad or your favorite word processor.  It
contains the descriptives for your variables separated by level.  I would
compare the descriptives from HLM to the ones in SPSS.  Once you identify
which variables differ, you can use the filter command to identify the cases
that are missing (select cases that fall out of the ranges listed in the MDM
statistics file).

    Also, what versions of HLM & SPSS are you using?

    I hope this helps.

        Best,
            Lisa



----- Original Message -----
From: "nschell" <[e-mail]>
To: <[e-mail]>
Sent: Monday, October 15, 2007 1:34 PM
Subject: HLM dropped cases


> Hello,
>
> I am importing data from SPSS into HLM.  I have three levels of data.  I
> have no missing data, the data is sorted correctly, and all the IDs match
> up.  However, when I import the data, HLM drops 19/301 cases at level 2.
> I
> imported levels 1&2 and 2&3 separately as two-level models to narrow down
> where the problem was, and it seems to be between levels 1&2, but I have
> no
> idea how to find out which cases it is dropping or why.  Does anyone have
> suggestions aside from missing data, mismatched IDs or sorting?  Does
> anyone
> know how to look at the data in HLM...just looking at the descriptives
> tells
> me I'm missing data but not which cases.  Or any suggestions on how to
> figure out which cases are being dropped?
>
> Thank you,
> Natalie
> --
> View this message in context:
> http://www.nabble.com/HLM-dropped-cases-tf4629052.html#a13217714
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
>
>>>> Error in line 4 of spssx-l.mailtpl: unknown formatting command <<<
> -> .................... <-
>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[parisec]

Hi all,

I'm running a mulitinomial logistic and my results show statistically significant interactions. I have plenty of cases so my results are pretty solid IMHO. The only annotated examples i can find (including Norusis' Advanced stats companion) are with interaction terms that are not significant. Easier, yes..but certainly not as interesting.


I have a DV (lets say SES for this example) that has 5 levels (1 though 5 with 5 being the ref cat)

IVs are Race: Black, Hispanic, Asian with White as the refcat
          Age:  <45, 46-70, 71+ with age 46-70 as refcat

I included the race x age interaction and came out with results that show sig interactions but i'm not quite sure how to interpret these. I understand that it means that the effect of race on SES depends on what age category people are in but the specific "as compared to" statements are elluding me with the 5 DVs.

Here are some examples:

For SES level 2:
        Age 71 * Black OR =.74
        Age 71 * Asian OR = 1.26
        Age <45 * Asian OR = .77

all of these are stat sig. with tight confidence intervals.

Is it correct to say:

Blacks aged 71+ are .74 times (26%)less likely than Blacks of other ages to have be in SES of 2 vs 5.
Asians aged 71+ are 1.26  times(26%) more likely than Asians of any other ages to be in SES of 2 vs 5.
Asians aged < 45 are 33% less likely than asians of other ages to be in SES 2 vs 5.

Is there a more correct way of interpreting these examples?

Thanks for any input you may have

Carol
Sacramento CA

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD